Zhu Jian-jun

Robustness and the robust estimate

The paper establishes a measure for robustness. With help of the measure the paper discusses the robust estimates under the stochastic error model and mean shift error model. The results show that under the measure and the stochastic error model the most reasonable robust estimate is the posterior variance approach proposed by Li Deren in 1983. Under the measure and the mean shift error model, the most robust estimate is the one with minimum mean squared error proposed by Zhu Jiangjun in 1991.

Event-based incremental updating of spatio-temporal database

Based on the relationship among the geographic events, spatial changes and the database operations, a new automatic (semi-automatic) incremental updating approach of spatio-temporal database (STDB) named as event-based incremental updating (E-BIU) is proposed in this paper. At first, the relationship among the events, spatial changes and the database operations is analyzed, then a total architecture of E-BIU implementation is designed, which includes an event queue, three managers and two sets of rules, each component is presented in detail. The process of the E-BIU of master STDB is described successively. An example of building’s incremental updating is given to illustrate this approach at the end. The result shows that E-BIU is an efficient automatic updating approach for master STDB.

Robust estimate with minimum mean squared error

Abstract The result of any robust estimate depends on its weight function. But the weight function and its parameters are determined personally. This paper establishes a new robust estimate which has a property of minimum mean squared error, all the parameters of which are determined according to statistics. Some numerical examples are given to show its feasibility.

Mathematical model research on landslide monitoring through GPS

Assuming that road slope and landslide object are rigid, the landslide’s moving displacement was drawn based on their geometry shapes and the physi-mechanical features of materials, and the dynamic model of landslide was also set up, then DDOD(double difference observation data) was combined with the deformed monitoring point and the carrier phase observation data on base point, which can be used to monitor the landslide’s deformation rule from horizontal, vertical and directional view simultaneously. Observing equation was set up, which sufficiently reflects the activities of landslide in entire directions. Filter model includes some information such as mechanical state and GPS observing data by Kalman filter.

Schreiber approach for GPS carrier phase ambiguity resolution

Differenced GPS carrier phase observations are usually used in GPS positioning to eliminate the various common GPS errors. A different approach is proposed in this paper where no differencing of the observations is required. In this method, the common GPS errors are modelled explicidy and eliminated by introducing the Schreiber method, which is used in traditional geodesy to eliminate the azimuth unknowns in observation equations. The paper first introduces the Schreiber method, and then discusses its application to the various GPS common-mode error models. It is shown that if only the real ambiguity solution is concerned, the estimate of the position parameters from the Schreiber approach is the same as that from the differencing approach. However, the variance of the ambiguity estimate from the Schreiber approach is smaller than that from the differencing approach. Numerical examples are given to show that the results are reliable.

SCGM(1,m)c model with time-lag and its application in deformation analysis

Originated in the early 1990s, SCGM(1,m)c model has enjoyed widespread application in the fields of urban planning, society economy prediction and modal control in recent years. However, none of these applications have taken account of time-lag effects in the modeling process. Aiming at overcoming the defect, the authors introduced time-lag items into SCGM(1,m)c model and developed a SCGM(1,m)c model with time-lag, then discusses in detail some principal problems in the model, such as parameters estimation, model verifying, model prediction, etc. The model was used on a real slope monitoring project and compared with the conventional SCGM(1,m)c model. The results show an improvement of average models precision from 1.321 to 0.238 and total average of relative prediction errors from 12.41% to 7.98% when the modeling data length ranges from 29 to 48 in the slope monitoring case.

A new dual-baseline polarimetric SAR interferometry for vegetation height inversion using complex least squares adjustment

Vegetation height is an important parameter in quantifying the terrestrial carbon cycle effectively. P band synthetic aperture radar (SAR) is good at monitoring forest areas, which is sensitive to the contributions from vegetation layer and ground. Random Volume over Ground (RVoG) model has been widely applied to forest height inversion. Corresponding single baseline or muti-baseline methods have been established, such as three-stage and nonlinear iteration methods, whose good performance have been test with differential wavelength and areas. However, all the methods do not pay much attention to the complex coherence errors, and take measure to reduce the errors. In this paper, a new dual-baseline polarimetric SAR interferometry vegetation height inversion method, dual-baseline complex least squares adjustment (DBCLSA), is proposed. The DBCLSA method behind the concept of complex least squares adjustment. The main idea of DBCLSA method is as follow. Firstly, the adjustment model is summarized as combined adjustment of complex real and imagine parts. And then, stochastic model is established based on Cramer-Rao bounds. After that, we show the linearization method and parameter solving method. At last, the obtained volume-only coherence values are used to vegetation height retrieval. The DBCLSA is validated on E-SAR P band data of BioSAR2008 in Sweden. The results of Cloude dual-baseline method are computed at the same time. Compared with ground true measurement data, the result of new approach is more accurate.

Improvement of GPS ambiguity resolution using prior height information. Part II: The method of using quasi observation

This paper deals with the method of using quasi observation. In the paper a simple algorithm is developed for the adjustment computation with quasi observation at first. And then the ability of quasi observation to improve ambiguity search technique is studied in detail. The robustness of the method is also discussed. A method to determine the weight of quasi observation is proposed. The results show that a prior height can be taken as a quasi observation and used together with GPS observations. It can strengthen residual tests, especially in situation where there are fewer satellites in the sky. It also can change structure of incorrect solutions, which will theoretically make less incorrect solutions left in search space. At last the field tests are carried out to show that the proposed method is effective. The success rate of ambiguity resolution in the four field tests is improved significantly.

The measure of mean cook distance under the contaminated error model

Redefines the mean Cook distance under the contaminated error model. With the help of the mean Cook distance, the paper discusses the relationship between local influence of small perturbations and high leverage case. A new robust method for the adjustment of geodetic networks is proposed. The suggested method, a generalization of the robust method with the minimum mean Cook distance, is more efficient than the others. The basic feature of the method is that the equivalent weight functions of the robust estimates are determined according to the principle of statistics.